@inproceedings{banasaz-nouri-etal-2026-novel,
title = "A Novel Synthetic Dataset for Few-Shot Legal Relation Extraction in {G}erman",
author = "Banasaz Nouri, Shiva and
Leitner, Elena and
Moreno-Schneider, Julian and
Rehm, Georg",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.830/",
doi = "10.63317/5oaz3mdtekn9",
pages = "10579--10591",
abstract = "The legal domain is particularly challenging for natural language processing due to the personal and confidential information it contains. Despite the significant advances of large language models (LLMs), applying them to relation extraction (RE) in legal texts remains challenging, not only because of the task{'}s linguistic and semantic complexity, but also due to privacy, compliance, and infrastructure constraints under regulations such as the EU AI Act. To address these challenges, we propose a novel synthetic dataset for German legal relation extraction, created using LLMs through a controlled, privacy-preserving, template-based pipeline. The dataset allows for reproducible and legally compliant experimentation. We benchmark it using two few-shot learning paradigms, a description-enhanced Model-Agnostic Meta-Learning (MAML) framework and Prototypical Networks with supervised contrastive loss and curriculum-aware prototype enrichment. Our results demonstrate that combining few-shot learning with structured semantic knowledge achieves robust and interpretable results, with the curriculum-aware Proto-Contrastive model reaching an F1-score of 99.83{\%}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="banasaz-nouri-etal-2026-novel">
<titleInfo>
<title>A Novel Synthetic Dataset for Few-Shot Legal Relation Extraction in German</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shiva</namePart>
<namePart type="family">Banasaz Nouri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elena</namePart>
<namePart type="family">Leitner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Julian</namePart>
<namePart type="family">Moreno-Schneider</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Georg</namePart>
<namePart type="family">Rehm</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The legal domain is particularly challenging for natural language processing due to the personal and confidential information it contains. Despite the significant advances of large language models (LLMs), applying them to relation extraction (RE) in legal texts remains challenging, not only because of the task’s linguistic and semantic complexity, but also due to privacy, compliance, and infrastructure constraints under regulations such as the EU AI Act. To address these challenges, we propose a novel synthetic dataset for German legal relation extraction, created using LLMs through a controlled, privacy-preserving, template-based pipeline. The dataset allows for reproducible and legally compliant experimentation. We benchmark it using two few-shot learning paradigms, a description-enhanced Model-Agnostic Meta-Learning (MAML) framework and Prototypical Networks with supervised contrastive loss and curriculum-aware prototype enrichment. Our results demonstrate that combining few-shot learning with structured semantic knowledge achieves robust and interpretable results, with the curriculum-aware Proto-Contrastive model reaching an F1-score of 99.83%.</abstract>
<identifier type="citekey">banasaz-nouri-etal-2026-novel</identifier>
<identifier type="doi">10.63317/5oaz3mdtekn9</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.830/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>10579</start>
<end>10591</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T A Novel Synthetic Dataset for Few-Shot Legal Relation Extraction in German
%A Banasaz Nouri, Shiva
%A Leitner, Elena
%A Moreno-Schneider, Julian
%A Rehm, Georg
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F banasaz-nouri-etal-2026-novel
%X The legal domain is particularly challenging for natural language processing due to the personal and confidential information it contains. Despite the significant advances of large language models (LLMs), applying them to relation extraction (RE) in legal texts remains challenging, not only because of the task’s linguistic and semantic complexity, but also due to privacy, compliance, and infrastructure constraints under regulations such as the EU AI Act. To address these challenges, we propose a novel synthetic dataset for German legal relation extraction, created using LLMs through a controlled, privacy-preserving, template-based pipeline. The dataset allows for reproducible and legally compliant experimentation. We benchmark it using two few-shot learning paradigms, a description-enhanced Model-Agnostic Meta-Learning (MAML) framework and Prototypical Networks with supervised contrastive loss and curriculum-aware prototype enrichment. Our results demonstrate that combining few-shot learning with structured semantic knowledge achieves robust and interpretable results, with the curriculum-aware Proto-Contrastive model reaching an F1-score of 99.83%.
%R 10.63317/5oaz3mdtekn9
%U https://aclanthology.org/2026.lrec-1.830/
%U https://doi.org/10.63317/5oaz3mdtekn9
%P 10579-10591
Markdown (Informal)
[A Novel Synthetic Dataset for Few-Shot Legal Relation Extraction in German](https://aclanthology.org/2026.lrec-1.830/) (Banasaz Nouri et al., LREC 2026)
ACL